Math-MCQ-Generator-v1

Model Description

This is a fine-tuned version of deepseek-ai/deepseek-math-7b-instruct specialized for generating high-quality mathematics multiple choice questions (MCQs). The model has been trained using QLoRA (Quantized Low-Rank Adaptation) to efficiently adapt the base model for educational content generation.

Capabilities

• Subject: Mathematics
• Question Types: Multiple Choice Questions (MCQs)
• Topics: Applications of Trigonometry, Conic Sections, and more
• Difficulty Levels: Easy, Medium, Hard
• Cognitive Skills: Recall, Direct Application, Pattern Recognition, Strategic Reasoning, Trap Aware

Training Information

• Base Model: deepseek-ai/deepseek-math-7b-instruct
• Training Method: QLoRA (4-bit quantization)
• Dataset Size: 1519 examples
• Training Epochs: 5
• Final Loss: ~0.20
• Training Date: 2025-09-03

Usage

Via Python API

from transformers import AutoTokenizer, AutoModelForCausalLM
from peft import PeftModel

# Load model
base_model = AutoModelForCausalLM.from_pretrained("deepseek-ai/deepseek-math-7b-instruct")
model = PeftModel.from_pretrained(base_model, "danxh/math-mcq-generator-v1")
tokenizer = AutoTokenizer.from_pretrained("danxh/math-mcq-generator-v1")

# Generate MCQ
prompt = '''### Instruction:
Generate a math MCQ similar in style to the provided examples.

### Input:
chapter: Applications of Trigonometry
topics: ['Heights and Distances']
Difficulty: medium
Cognitive Skill: direct_application

### Response:
'''

inputs = tokenizer(prompt, return_tensors="pt")
outputs = model.generate(**inputs, max_new_tokens=300, temperature=0.7)
response = tokenizer.decode(outputs[0], skip_special_tokens=True)

Performance

The model demonstrates strong performance in generating contextually appropriate mathematics MCQs with:

• Proper question formatting
• Relevant multiple choice options
• Appropriate difficulty scaling
• Subject-matter accuracy

License

MIT License - Feel free to use, modify, and distribute.